Training and evaluation code for the paper "Headless Language Models: Learning without Predicting with Contrastive Weight Tying" (https://arxiv.org/abs/2309.08351) - View it on GitHub